I have this values file =>

zookeeper:
    podManagementPolicy: OrderedReady
    securityContext:
      runAsGroup: 0
      runAsUser: 0
    configData:
      PULSAR_MEM: >
        -Xms64m -Xmx128m
      PULSAR_GC: >
        -XX:+UseG1GC
        -XX:MaxGCPauseMillis=10
        -Dcom.sun.management.jmxremote
        -Djute.maxbuffer=10485760
        -XX:+ParallelRefProcEnabled
        -XX:+UnlockExperimentalVMOptions
        -XX:+DoEscapeAnalysis
        -XX:+DisableExplicitGC
        -XX:+PerfDisableSharedMem
  bookkeeper:
    securityContext:
      runAsGroup: 0
      runAsUser: 0
    configData:
      PULSAR_MEM: >
        -Xms128m
        -Xmx1g
        -XX:MaxDirectMemorySize=1g
      PULSAR_GC: >
        -XX:+UseG1GC
        -XX:MaxGCPauseMillis=10
        -XX:+ParallelRefProcEnabled
        -XX:+UnlockExperimentalVMOptions
        -XX:+DoEscapeAnalysis
        -XX:ParallelGCThreads=4
        -XX:ConcGCThreads=4
        -XX:G1NewSizePercent=50
        -XX:+DisableExplicitGC
        -XX:-ResizePLAB
        -XX:+ExitOnOutOfMemoryError
        -XX:+PerfDisableSharedMem
        -verbosegc
      dbStorage_writeCacheMaxSizeMb: "32"
      dbStorage_readAheadCacheMaxSizeMb: "32"
      dbStorage_rocksDB_writeBufferSizeMB: "8"
      dbStorage_rocksDB_blockCacheSize: "8388608"
  broker:
    replicaCount: 4
    configData:
      PULSAR_MEM: >
        -Xms1024m -Xmx1024m -XX:MaxDirectMemorySize=256m
      PULSAR_GC: >
        -XX:+UseG1GC
        -XX:MaxGCPauseMillis=10
        -Dio.netty.leakDetectionLevel=disabled
        -Dio.netty.recycler.linkCapacity=1024
        -XX:+ParallelRefProcEnabled
        -XX:+UnlockExperimentalVMOptions
        -XX:+DoEscapeAnalysis
        -XX:ParallelGCThreads=4
        -XX:ConcGCThreads=4
        -XX:G1NewSizePercent=50
        -XX:+DisableExplicitGC
        -XX:-ResizePLAB
        -XX:+ExitOnOutOfMemoryError
        -XX:+PerfDisableSharedMem
      managedLedgerDefaultEnsembleSize: "2"
      managedLedgerDefaultWriteQuorum: "2"
      managedLedgerDefaultAckQuorum: "2"
      allowAutoTopicCreationType: "partitioned"
  proxy:
    replicaCount: 3
    restartPodsOnConfigMapChange: false
    configData:
      PULSAR_MEM: >
        -Xms64m -Xmx64m -XX:MaxDirectMemorySize=64m
      PULSAR_GC: >
        -XX:+UseG1GC
        -XX:MaxGCPauseMillis=10
        -Dio.netty.leakDetectionLevel=disabled
        -Dio.netty.recycler.linkCapacity=1024
        -XX:+ParallelRefProcEnabled
        -XX:+UnlockExperimentalVMOptions
        -XX:+DoEscapeAnalysis
        -XX:ParallelGCThreads=4
        -XX:ConcGCThreads=4
        -XX:G1NewSizePercent=50
        -XX:+DisableExplicitGC
        -XX:-ResizePLAB
        -XX:+ExitOnOutOfMemoryError
        -XX:+PerfDisableSharedMem
    service:
      type: NodePort

Did i miss something in my configuration ? is the rack placement policies the default data placement ?

I didn’t see the replica count for the Bookies, but based on your original statement, I assume it is 3?

Can you confirm that all bookies are running and that the brokers can communicate with them?

yes all Bookies are running and brokers can communicate with them

AZ deployment ?

AZ = availability zones. Different zones in the same region.

Given the uneven distribution of the data across the bookies, my guess is that there were some network connectivity issues with the bookies that have less data. This would cause the segments they were hosting to be closed prematurely, thus containing less data. This would also result in these bookies not getting selected to host new segments because the brokers couldn’t connect to them.

Even if these connectivity issues were transient and are no longer present, the skewed data distribution would remain. If the data distribution continues to be skewed, e.g., the data volume growth in nominal terms is still skewed to one bookie, then that would indicate that the issue remains.

It can be a network connectivity when it is on the same k8s cluster ?

all Bookies are on the same kubernetes cluster

For now, track the data distribution in the future. If the skew continues, then you have an ongoing issue. If not, it was a transient issue. The good news is that BK should be able to handle unbalanced data distribution just fine.

another important point that if I sum my tenant data on Pulsar Manager i have 266GB x 2 because its duplicated but when i sum my persistence between my bookies pvc it's 990 GB

IIRC, K8s PVC represents the total capacity you have allocated from the underlying storage class, i.e., “claimed”, and doesn’t represent the total capacity utilized.